The Realist approach offers a way to understand the complexity of social interventions by seeking to understand what works, for whom, and in what contexts. Based on a review of the scientific literature and a reflective analysis of our experiences, we present some challenges and reflections related to the use of this evaluative innovation. While the Realist approach appears intellectually appealing, constructing the "Context-Mechanism-Effect" configurations it suggests proves to be a delicate task. The difficulties of finding information on contexts or effects, distinguishing context elements from mechanism elements, and adopting the appropriate level of analysis all pose significant challenges in the use of the Realist approach. [author’s abstract]